[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <c5a521e0-26b1-b1d6-7f7d-00aa9b4b1e0e@suse.com>
Date: Wed, 15 Jun 2022 13:39:01 +0200
From: Juergen Gross <jgross@...e.com>
To: Christoph Hellwig <hch@...radead.org>
Cc: xen-devel@...ts.xenproject.org, linux-doc@...r.kernel.org,
linux-kernel@...r.kernel.org, Jonathan Corbet <corbet@....net>,
Stefano Stabellini <sstabellini@...nel.org>,
Oleksandr Tyshchenko <oleksandr_tyshchenko@...m.com>
Subject: Re: [PATCH] xen: don't require virtio with grants for non-PV guests
On 15.06.22 13:28, Christoph Hellwig wrote:
> On Wed, Jun 15, 2022 at 01:26:27PM +0200, Juergen Gross wrote:
>> On 15.06.22 13:24, Christoph Hellwig wrote:
>>> I don't think this command line mess is a good idea. Instead the
>>> brand new grant devices need to be properly advertised in the device
>>> tree, and any device that isn't a grant device doesn't need
>>> VIRTIO_F_ACCESS_PLATFORM. No need for a command line or user
>>> intervention.
>>
>> And on x86?
>
> ACPI tables or whatever mechanism you otherwise use to advertise grant
> based DMA. But that is irrelevant for now, as the code in mainline
> only supports DT on arm/arm64 anyay.
No, it doesn't. I'm working on a qemu patch series enabling the qemu
based backends to support grants with virtio. The code is working fine
on x86, too (apart from the fact that the backends aren't ready yet).
I'd be fine to just drop the command line parameter, but I think a
guest admin should be able to specify that he doesn't want the backend
to be able to map arbitrary memory of the guest in order to add
another line of defense.
Juergen
Download attachment "OpenPGP_0xB0DE9DD628BF132F.asc" of type "application/pgp-keys" (3099 bytes)
Download attachment "OpenPGP_signature" of type "application/pgp-signature" (496 bytes)
Powered by blists - more mailing lists